Genuine question: What do people feel they understand after going through this? If we wrote out every matrix multiplication for linear regression, would we say we've truly grasped it? Is the takeaway about implementation mechanics, or does it build deeper intuition?<p>I see the value in visualization, but what’s the real educational gain beyond that?<p>(I'm an ML person/mathematician, but I haven’t lectured in over a decade—maybe I just don’t get it. I tend to prefer tutorials that build connections to known ideas, so this might be a blind spot for me.)
On the related topic to teach AI fundamentals and intuition of the inner workings, there is a wonderful material named "AI Unplugged"[1] to perform activities with pen/pencils, cards, etc. in a game manner.<p>I've been using this material on several occasions with various audiences not familiar with the field of AI/ML (kids and grown-ups), and each time people seem to have been enjoying it and earning a bit of understanding of the modern world they live in.<p>[1] <a href="http://www.aiunplugged.org" rel="nofollow">http://www.aiunplugged.org</a>
You can even build GPT-2 in Excel: <a href="https://www.youtube.com/watch?v=yjD2n_e9E3w&ab_channel=Spreadsheetsareallyouneed" rel="nofollow">https://www.youtube.com/watch?v=yjD2n_e9E3w&ab_channel=Sprea...</a>
One can also argue this can be done within db such as postgres have everything store procedures/functions/triggers. In some ways DBs are just Spreadsheets with really robust querying languages.
Similar HN thread: <a href="https://news.ycombinator.com/item?id=42967173">https://news.ycombinator.com/item?id=42967173</a>
See also:<p>(Classical) Computer vision basics in Excel, using just formulas: <a href="https://github.com/amzn/computer-vision-basics-in-microsoft-excel">https://github.com/amzn/computer-vision-basics-in-microsoft-...</a><p>Original HN Submission: <a href="https://news.ycombinator.com/item?id=22357374">https://news.ycombinator.com/item?id=22357374</a><p>PS: I am the primary author.